MIMIC-IV-Echo: An Echocardiography Dataset Integrated with the MIMIC-IV Ecosystem

Abstract

MIMIC-IV-Echo is a publicly available dataset of echocardiographic imaging and structured quantitative measurements linked to the MIMIC-IV clinical database. The module includes more than 7,000 transthoracic echocardiogram studies in DICOM format acquired between 2017 and 2019, accompanied by over 200,000 structured measurement studies spanning transthoracic, transesophageal, and stress echocardiography from 2008 through 2022. Imaging files retain their original multi-frame structure and acquisition characteristics, while structured measurements provide clinician-reviewed quantitative parameters recorded during routine clinical interpretation. All records are deidentified and linked to the broader MIMIC-IV ecosystem through shared identifiers and date shifting, enabling integration with hospital, ICU, emergency department, laboratory, medication, and clinical note data. By providing both raw echocardiographic imaging and associated quantitative measurements, MIMIC-IV-Echo supports a wide range of research applications in cardiovascular imaging, multimodal learning, phenotyping, and clinical outcome modeling. The dataset is freely available on PhysioNet.